Untangling Gloxinieae (Gesneriaceae). II. Reconstructing Biogeographic Patterns and Estimating Divergence Times Among New World Continental and Island Lineages

Abstract

Gesneriaceae tribe Gloxinieae is a diverse clade of approximately 19 genera and 215 species. As with many tropical lineages, patterns and timing of diversification are poorly understood. This is a particular difficulty in groups such as the Gesneriaceae that have no fossil record. Here we explore maximum likelihood and Bayesian inference of phylogenetic relationships in the tribe based on nuclear, chloroplast, and morphological data sets, use Fitch parsimony optimization (FPO) and dispersal vicariance (DIVA) analyses to explore biogeographic patterns in the Gesnerioideae, and use penalized likelihood calibrated by geological events in the Caribbean and South America to explore timing of movement of lineages among Caribbean, Central American, and South American land masses and islands. Likelihood and Bayesian analyses increase support of previous hypotheses of relationships using parsimony and provide additional resolution in some parts of the phylogeny. FPO and DIVA analyses suggest that the most likely scenario for movement among Central American, Caribbean, and South American areas was either an early dispersal to Central America and the Caribbean prior to diversification of the Gloxinieae clade with subsequent back dispersal to South America, or the ancestor of the Gloxinieae had a broad distribution across Central America and Andean/western South America. Estimations of the timing of movement of these lineages among these land masses suggests that the Greater Antilles/Aves Ridge landbridge likely played a role in dispersal events and that the Gloxinieae/Gesnerieae lineage likely arrived in the Central America/Caribbean zone at least 26 million years ago.

Full text

Untangling Gloxinieae (Gesneriaceae). II. Reconstructing Biogeographic Patterns and
Estimating Divergence Times Among New World Continental and Island Lineages
Eric H. Roalson,
1,4
Laurence E. Skog,
2
and Elizabeth A. Zimmer
2,3
1
School of Biological Sciences and Center for Integrated Biotechnology, Washington State University,
Pullman, Washington 99164-4236 U.S.A.
2
Department of Botany, National Museum of Natural History, Smithsonian Institution,
Washington, District of Columbia 20560-0166 U.S.A.
3
Laboratories of Analytical Biology, National Museum of Natural History, Smithsonian Institution,
Suitland, Maryland 20746 U.S.A.
4
Author for correspondence (roalson@mail.wsu.edu)
Communicating Editor: Lena Struwe
Abstract—Gesneriaceae tribe Gloxinieae is a diverse clade of approximately 19 genera and 215 species. As with many tropical lineages,
patterns and timing of diversification are poorly understood. This is a particular difficulty in groups such as the Gesneriaceae that have no
fossil record. Here we explore maximum likelihood and Bayesian inference of phylogenetic relationships in the tribe based on nuclear,
chloroplast, and morphological data sets, use Fitch parsimony optimization (FPO) and dispersal vicariance (DIVA) analyses to explore
biogeographic patterns in the Gesnerioideae, and use penalized likelihood calibrated by geological events in the Caribbean and South
America to explore timing of movement of lineages among Caribbean, Central American, and South American land masses and islands.
Likelihood and Bayesian analyses increase support of previous hypotheses of relationships using parsimony and provide additional reso-
lution in some parts of the phylogeny. FPO and DIVA analyses suggest that the most likely scenario for movement among Central American,
Caribbean, and South American areas was either an early dispersal to Central America and the Caribbean prior to diversification of the
Gloxinieae clade with subsequent back dispersal to South America, or the ancestor of the Gloxinieae had a broad distribution across Central
America and Andean/western South America. Estimations of the timing of movement of these lineages among these land masses suggests
that the Greater Antilles/Aves Ridge landbridge likely played a role in dispersal events and that the Gloxinieae/Gesnerieae lineage likely
arrived in the Central America/Caribbean zone at least 26 million years ago.
Keywords—Bayesian inference, biogeography, Caribbean, divergence dating, Gesneriaceae, Gloxinieae.
Gesneriaceae tribe Gloxinieae is a morphologically diverse
clade currently considered to include approximately 19 gen-
era and 215 species (Roalson et al. 2005a). This tribe forms a
well-defined and exclusively New World clade of subfamily
Gesnerioideae (Zimmer et al. 2002; Roalson et al. 2005a). His-
torically, this group has been extremely problematic to clas-
sify due to a complex pattern of diversification and conver-
gence in floral and vegetative forms. This has led to numer-
ous reorganizations of generic boundaries in the tribe (e.g.
Wiehler 1983) based on traditional systematic lines of evi-
dence. Only recently have reasonably consistent and resolved
phylogenetic patterns of this tribe begun to emerge (Zimmer
et al. 2002; Roalson et al. 2003, 2005a). While this tribe has
been included in several molecular phylogenetic studies over
the past 10 yr (Smith 1996; Smith and Carroll 1997; Smith et
al. 1997a, b; Smith and Atkinson 1998; Smith 2000c, 2001;
Smith et al. 2004a, b), these studies did not provide consistent
nor well-supported branching patterns in the Gloxinieae (for
a more complete discussion see Zimmer et al. 2002).
While considerable attention has been paid to morphology,
classification, and phylogeny in the Gesneriaceae, little atten-
tion has been paid to the historical biogeography of the fam-
ily, or of any major clades within the family. To date, only
Old World members of the family have been discussed in this
context (Möller and Cronk 1997; Burtt 1998; Denduangbori-
pant and Cronk 2000; Atkins et al. 2001; Denduangboripant
et al. 2001; Mendum et al. 2001; Möller and Cronk 2001). As
with many tropical groups, neotropical gesneriads are very
poorly understood with regards to biogeographic history,
with effectively no studies exploring potential explanations
of distribution patterns in the family in the New World.
Part of the difficulty of determining biogeographic pat-
terns in the Gesneriaceae is the lack of any fossil pollen or
macrofossils. Many other lineages of the Lamiales also have
poor fossil records, and the earliest fossil pollen records from
families including the Lamiaceae, Acanthaceae, Orobancha-
ceae, and Plantaginaceae do not show up until the lower
Miocene (∼15–25 million years ago [MYA]) at the earliest
(Muller 1981; Wiehler 1983). There seems to be a very differ-
ent inference of the timing of diversification of Lamiales lin-
eages based upon these pollen records as opposed to those
inferences derived from our current evidence of phylogenetic
relationships. The Gesneriaceae and Oleaceae appear to be
two of the earliest diverging lineages of the Lamiales (Oxel-
man et al. 1999; Olmstead et al. 2000), but the pollen record
for the Gesneriaceae appears to be completely absent and
Oleaceae pollen has been dated as early as the Oligocene
(∼25–35 MYA; Muller 1981). The Bignoniaceae, which does
not diverge from its sister taxa (likely some portion of the
Verbenaceae, Scrophulariaceae, or Buddlejaceae; Oxelman et
al. 1999; Olmstead et al. 2000) until substantially later than
the origin of the Oleaceae or Gesneriaceae lineages has pollen
dated from the middle Eocene, 15–20 MYA prior to the ear-
liest record of the Oleaceae. Similarly, macrofossils of Acan-
thus (Acanthaceae) have been found as early as the Eocene,
despite the more recent divergence of the Acanthaceae from
other Lamiales lineages (Raven and Axelrod 1974). These
findings suggest that either the pollen and/or macrofossils
from the Lamiales need to be reassessed with regards to cur-
rent phylogenetic/classification hypotheses, or the fossil rec-
ord is exceedingly misrepresentative of the true antiquity of
several of these lineages, and particularly the Gesneriaceae.
Comparison of Gesneriaceae diversity and biogeography
in relation to plate tectonics led Raven and Axelrod (1974) to
suggest that, given the distribution of subfamily Gesnerioi-
deae in South America and the distribution of the Cyrtan-
droideae in Africa and Australasia, the family likely dates to
a time when South America and Africa were in close physical
Systematic Botany (2008), 33(1): pp. 159–175
© Copyright 2008 by the American Society of Plant Taxonomists
159

proximity, with the divergence of these lineages associated
with the separating of these continents (continental vicari-
ance). This would place the origins of the family prior to the
Mesozoic/Cenozoic boundary at 65 MYA, as these two con-
tinents were reasonably separated by this time (McLoughlin
2001). Alternatively, the origin of the subfamilies may have
postdated the division of South America and Africa and their
distribution may be the result of more recent long-distance
dispersal between these regions.
Large-scale estimation of nodal ages of major angiosperm
clades has suggested a wide range of potential dates for the
origin of the Lamiales clade, including as early as 71–74 MYA
(Wikström et al. 2001) or as late as 44.25 MYA (Magallón and
Sanderson 2001). Bremer et al. (2004) estimated a date for the
age of the crown group Gesneriaceae at 71 MYA based on
analyses of divergence dates across the Asterids. It should be
noted, however, that current methods of inferring ages at
phylogenetic nodes can be substantially influenced by phy-
logenetic uncertainty, substantial noise, and lineage affects,
creating exceptionally large confidence intervals around age
estimates (Sanderson and Doyle 2001).
Similar difficulty lies in understanding the biogeographic
patterns within the Gesnerioideae lineage and the patterns of
diversity found in different regions of South America, Cen-
tral America, and the Caribbean. Some of this comes from
incomplete understanding of the diversity and distributions
of these lineages, but is also associated with the previous
dependence on “centers of diversity”to explain biogeo-
graphic patterns (Gentry 1982; Kvist and Skog 1992, 1996;
Skog and Kvist 2000; Weigend and Förther 2002; Perret et al.
2003). While there are potential problems with the applica-
tion of the “centers of diversity”method to understanding
biogeography, it should be noted that these authors are
among very few Gesneriaceae researchers who have even
made an attempt to address biogeographic patterns, and the
lack of discussion of biogeographic issues in major works on
Gesneriaceae (Wiehler 1976, 1983; Smith and Carroll 1997;
Smith and Atkinson 1998; Smith 2000a, 2000b, 2001) under-
scores this problem.
It is clear that the largest diversity of New World Gesne-
riaceae is found in northwestern South American mideleva-
tion rainforests and cloud forests (Gentry 1982; Kvist and
Skog 1992, 1996; Skog and Kvist 2000), although some lin-
eages are quite diverse in other regions (e.g. Sinningia in
southeastern Brazil, Perret et al. 2003; Achimenes in Mexico,
Ramírez Roa 1987; Gesneria in the Greater Antilles, Skog
1976). The traditional view that movement of lineages into
other areas of the Neotropics was primarily unidirectional
from this center of diversity is likely overly simplistic. How-
ever, alternative hypotheses of lineage movement have not
been presented and the Gesneriaceae literature lacks any ref-
erence to biogeographic patterns outside of mapping loca-
tions of small groups under study (e.g. Kvist and Skog 1996).
The plate tectonic movements and origins of current land-
forms in northern South America, Central America, and the
Caribbean Islands are exceedingly complex (Raven and Ax-
elrod 1974; Coney 1982; Burnham and Graham 1999; Itur-
ralde-Vinent and MacPhee 1999). While the timing of the
Panamanian land bridge between Central America and Co-
lombia is solidly placed between 3.0 and 3.5 MYA (Stehli and
Webb 1985; Knowlton et al. 1993), there is some question
about the presence and persistence of other land masses that
might have facilitated movement of angiosperms among the
larger land masses. Particularly of interest are the island
chains creating potential links between northwestern South
America and Central America (Raven and Axelrod 1974; Itur-
ralde-Vinent and MacPhee 1999), and the Lesser Antillean
island chains and Greater Antilles/Aves Ridge landbridge
(GAARLANDIA) thought to connect northeastern South
America to the Greater Antilles and into close proximity to
the North American land mass as early as 35 MYA (Iturralde-
Vinent and MacPhee 1999). The potential role of these
bridges and island chains in movement of the Gesneriaceae
to the Caribbean and North and Central America is unclear,
and, as discussed by Raven and Axelrod (1974), different
lineages of Gesneriaceae likely arrived in these areas at dif-
ferent times, and therefore likely were differentially affected
by the various possible paths. Some lineages, such as the
Gesnerieae tribe, are quite prominent in the West Indies sup-
porting “the notion of a Paleogene arrival”(Raven and Ax-
elrod 1974), while Columnea with 11 endemic species on Ja-
maica might provide evidence for their arrival “early in Neo-
gene time”(Raven and Axelrod 1974). These estimated
arrival times were loosely based on the numbers of island
endemics in these genera. Colonization of Caribbean islands
prior to Late Eocene (∼35 MYA) is not generally considered
plausible as there were no permanent land masses prior to
this epoch (Iturralde-Vinent and MacPhee 1999).
In order to start to understand biogeographic patterns and
potential timing and mechanisms of dispersal, geographic
distributions of lineages need to be placed in a phylogenetic
context. Here we explore biogeographic patterns in the Ges-
neriaceae and particularly of tribe Gloxinieae, using phylo-
genetic hypotheses derived from maximum likelihood and
Bayesian inference analyses of a variety of molecular and
morphological data sets. The Gloxinieae are particularly ap-
propriate for dating the movement among the New World
land masses because of the presence of diverse lineages
within this tribe largely restricted to one or a few of the
geographic areas.
Likelihood and Bayesian phylogenetic hypotheses provide
a framework for the exploration of historical biogeographic
patterns with Fitch parsimony optimization and dispersal-
vicariance analyses. Furthermore, the timing of biogeo-
graphic movements in the Gesnerioideae are addressed by
dating nodes using plate tectonic and land bridge geologic
time estimates and semiparametric rate smoothing by penal-
ized likelihood methods. These results are discussed in the
framework of previous biogeographic hypotheses and the
potential for future dating of ancestral nodes in the Gloxin-
ieae and Gesneriaceae as a whole is explored.
MATERIALS AND METHODS
Taxon Sampling, Gesnerioideae—Samples included in this study are
those used in previous phylogenetic analysis of relationships in the Ges-
nerioideae (Zimmer et al. 2002; Roalson et al. 2005a). This includes 34
genera and 58 species in the Gesnerioideae and uses members of the
Cyrtandroideae (Aeschynanthus hildebrandii and Streptocarpus primuli-
folius) as outgroups (Zimmer et al. 2002). The three data sets include
nrDNA internal transcribed spacer (ITS) sequences, cpDNA trnL intron
and trnL–F intergenic spacer (trnL–F) sequences, and cpDNA trnE–T in-
tergenic spacer sequences (trnE–T). Details of these datasets have been
previously enumerated, and will not be repeated here. All samples used
in this study are vouchered or presented in Zimmer et al. (2002) and
Roalson et al. (2005a) and listed in Appendix 1. Nomenclature follows
Roalson et al. (2005b). The data matrix and maximum likelihood tree have
been deposited in TreeBASE (study number S1794).
160 SYSTEMATIC BOTANY [Volume 33

Taxon Sampling, Gloxinieae—Samples included in this study are those
used in previous phylogenetic analysis of relationships in the Gloxinieae
(Roalson et al. 2005a) and Gesnerioideae (Zimmer et al. 2002). Gloxinieae
analyses include 17 genera and 54 ingroup species with members of the
Gesnerieae (six species in four genera) as outgroups (Roalson et al.
2005a). The three data sets include nrDNA internal transcribed spacer
(ITS) sequences, cpDNA trnL intron and trnL–Fintergenic spacer (trnL–F)
sequences, and a morphological dataset. Details of these datasets have
been previously enumerated, and will not be repeated here. All samples
used in this study are vouchered or presented in Zimmer et al. (2002) and
Roalson et al. (2005a) and listed in Appendix 1. Nomenclature changes
suggested in Roalson et al. (2005b) are followed here and differ from the
generic circumscriptions used in Zimmer et al. (2002) and Roalson et al.
(2005a). The data matrix and maximum likelihood tree have been depos-
ited in TreeBASE (S1794).
Maximum Likelihood Phylogeny, Gesnerioideae—Maximum likeli-
hood (ML) analysis of the combined ITS/trnL–F/trnE–Tdataset was per-
formed using PAUP* 4.0b10 (Swofford 2001). Heuristic searches were
employed (TBR branch swapping). ML analyses employed the general
time reversable (GTR; Rodríguez et al. 1990) model with proportion of
invariant sites (I) and gamma shape (G) parameters and empirical base
frequencies (six substitution types: A/C, 0.8489; A/G, 1.5790; A/T,
0.4177; C/G, 0.8349; C/T, 3.1447; G/T, 1.0000; I = 0.2954; G = 0.4704; A,
0.3151; C, 0.1904; G, 0.1931; T, 0.3014). This model was chosen based on
the results of analysis using DT_ModSel (Minin et al. 2003). DT-ModSel
examines the fit of various substitution models to the data set using the
Bayesian information criterion and additionally incorporates relative
branch-length error estimates in a decision theory framework (Minin et
al. 2003).
Maximum Likelihood Phylogeny and Tests of Alternative Topologies,
Gloxinieae—Maximum likelihood (ML) analysis of the combined ITS/
trnL–Fdataset was performed using PAUP* 4.0b10 (Swofford 2001). Heu-
ristic searches were employed (TBR branch swapping). ML analyses em-
ployed the Tamura and Nei (TrN; Tamura and Nei 1993) model with
proportion of invariant sites (I) and gamma shape (G) parameters and
empirical base frequencies (six substitution types: A/C, 1.0000; A/G,
2.5060; A/T, 1.0000; C/G, 1.0000; C/T, 4.3089; G/T, 1.0000; I = 0.4964; G
= 0.6625; A, 0.3018; C, 0.1973; G, 0.2018; T, 0.2991). This model was chosen
based on the results of analysis using DT_ModSel (see description above;
Minin et al. 2003). One alternative topology was tested using the Shimo-
daira-Hasegawa (SH) test (Shimodaira and Hasegawa 1999) in a maxi-
mum likelihood framework employing constraint options implemented
in PAUP*: the topology forced monophyly of a clade containing all Cen-
tral American genera (Achimenes,Eucodonia,Moussonia,Niphaea,Smithi-
antha, and Solenophora). The SH analysis was run with 10000 RELL (Re-
sampling of Estimated Log Likelihoods) bootstraps (one-tailed).
Bayesian Inference Analysis of Three Datasets, Gloxinieae—Bayesian
inference analyses were performed using MrBayes v.3.0 (Huelsenbeck
and Ronquist 2001). Three partitions were set to correspond with the ITS,
trnL–F, and morphological datasets. The parameters for each dataset
were allowed to vary independently (“unlinked”). Priors for the two
molecular datasets included a model with six substitution types, rates
following a gamma distribution (four categories), and allowing a propor-
tion of invariant sites in the ITS model. These models were chosen based
on the results of analyses using DT-ModSel (see description above; Minin
et al. 2003). Priors for the morphological dataset were set at the default
condition of all character changes equally probable and equally weighted.
One hundred million generations were run with four Markov Chain
Monte Carlo (MCMC) chains, and a tree was saved every 100 generations.
The trees from the MrBayes analysis were loaded into PAUP*4.0b10
(Swofford 2001), discarding the trees sampled during the “burnin”of the
chain (Huelsenbeck and Ronquist 2001; the first 20,000,000 generations or
the first 200,001 trees) to only include trees after stationarity was reached.
Examination of likelihood plots suggest that stationarity was reached
within the designated burnin (data not shown). Additionally, multiple
independent runs (five) starting from different random trees were con-
ducted to determine if convergence and mixing had occurred. A majority
rule consensus tree was made, showing nodes with a posterior probabil-
ity of 50% or more. Majority rule consensus trees of the trees sampled in
Bayesian inference analyses yielded probabilities that the clades are
monophyletic (Lewis 2001). Additionally, five trees were randomly cho-
sen (chosen using a random number generator at http://random.org)
from the posterior probability distribution of trees for use in the character
and biogeography reconstructions described below. These five trees rep-
resent different topologies from the posterior probability distributionof
trees and therefore represent plausible topologies, although some nodes
on some trees might have a lower probability than others. Since biogeo-
graphic hypotheses are being explored across several plausible topolo-
gies, hypotheses on the evolutionary dynamics in the group will be more
robust since they are not reliant on a single topology being correct at all
nodes.
Reconstruction of Biogeographic Patterns—The historical biogeo-
graphic patterns in the Gesnerioideae and Gloxinieae datasets were in-
ferred with standard Fitch parsimony optimization (FPO; accelerated
transformation; as implemented in Mesquite 1.12; Maddison and Mad-
dison 2006) and dispersal-vicariance analysis (DIVA; Ronquist 1996,
1997). The FPO method assumes that geographic distributions are the
result of dispersal events. Polymorphic nodes are therefore restricted to
terminal nodes and ancestral states are calculated by minimizing the
number of character state changes on the tree. For these analyses geo-
graphic distribution was coded as a single multistate character based on
distributions as described in Appendices 2 and 3 and analyzed using
Mesquite 1.12 (Maddison and Maddison 2006). DIVA analysis, alterna-
tively, assumes that geographic distributions can be the result of dis-
persal, extinction, and vicariance events. Polymorphic states are not re-
stricted to terminals and ancestral states are calculated by minimizing the
number of dispersal events necessary to explain the distribution pattern.
Geographic distributions were coded as separate binary characters for
presence/absence of each species in each geographic region based on
distributions as described in Appendices 2 and 3 and the analysis was
performed using DIVA 1.1a (Ronquist 1996).
The four areas described for FPO and DIVA analyses of the Gloxinieae
were the Caribbean, Central America, Andean/western South America,
and southern Brazil. Gesnerioideae analyses also included Old World
and Guianas geographic areas. DIVA optimizations were conducted with
either an unrestricted maximum number of areas assigned to each node
or with the maximum number of areas restricted to two. This follows the
reasoning that species of the Gesnerioideae are generally not found in
more than two of the geographic regions described here, and that forcing
internal nodes to not include all geographic regions may tell us more
about patterns of dispersal and vicariance than would suggesting that
ancestors were extremely widespread in the New World. Both FPO and
DIVA analyses were run on all five of the trees randomly chosen from the
posterior probability distribution of trees resulting from the Bayesian
analysis, however, the primary differences in ancestral state reconstruc-
tions revolve around the paraphyly of the Gloxinieae Central American
clades in relation to the Gloxinieae South American clade or the mono-
phyly of the Central American clade. Therefore, a representative tree of
each type is presented here, but it should be noted that other slightly
different topologies had identical reconstructions of the nodes of interest
(data not shown).
Estimating Absolute Divergence Times of Gesnerioideae Clades—
There are many methods now available for phylogenetic dating (for re-
views see Benton and Ayala 2003; Bromham and Penny 2003). Here we
used the r8s program (Sanderson 2004) implementing semiparametric
rate smoothing by penalized likelihood and the truncated Newton algo-
rithm (Sanderson 2002) which uses a likelihood approach with a rough-
ness penalty that prevents too much variation in rates. The roughness
penalty is specified by a smoothing parameter derived from cross-
validation analyses. Cross validation analyses (not shown) resulted in a
smoothing parameter of 160.
The topology and branch lengths from the Gesnerioideae ML analysis
were used as the starting point for estimating absolute divergence dates.
Due to the r8s restriction that all nodes be resolved, unresolved branches
from the ML analysis were resolved with very short branches using the
topology found in Fig. 3, based on previous analyses. Five different con-
straint sets were used to assess divergence times on the ML topology.
These constraint sets set maximum or maximum and minimum ages at
two or three tree nodes and are referred to as “56/52,”“56/32,”“56/8,”
“56/52/8,”and “56/32/8”(Table 1). All analyses included a maximum
age at node 56 of 71 MYBP based on estimates of the stem age of the
Gesneriaceae (Bremer et al. 2004). Bremer et al. (2004) included Peltanthera
within their circumscription of Gesneriaceae for this estimate, which is a
circumscription that other Gesneriaceae researchers disagree with (Wang
et al. 2004). However, this date is used as a maximum for the Gesneria-
ceae stem age in this study, and is therefore appropriately used here even
if Peltanthera is not considered part of Gesneriaceae. Geological evidence
suggests that GAARLANDIA and persistent Caribbean land masses did
not predate 35 MYBP and that the GAARLANDIA land bridge had been
sundered by 25 MYBP. In order to test the possible association of the
Gesnerieae + Gloxinieae stem and crown nodes with this land bridge, 25
and 35 MYBP were placed as minimum and maximum ages on nodes 52
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 161

and 32. By approximately 16 MYBP, a southern Central America tectonic
block narrowed the gap between the North American and South Ameri-
can land masses. In order to explore this land mass’role in facilitating
migration of the Gloxinieae back to South America, node 8 was con-
strained with a maximum age of 16 MYBP. Each of these three constraints
was analyzed with the family age constraint individually. Combined con-
straint of nodes 56, 52, and 8 (but with a maximum constraint only on
node 52), and nodes 56, 32, and 8 (but with a maximum constraint only
on node 32) were each analyzed. All land mass ages were taken from
Iturralde-Vinent and MacPhee (1999).
Confidence intervals were calculated by creating 100 bootstrap data
matrices of the Gesnerioideae gene matrix using the SEQBOOT program
in Felsenstein’s (2004) PHYLIP package. These replicate data sets were
used to estimate branch lengths on the ML topology and these phylo-
grams were then analyzed in r8s to create a 95% confidence interval of the
age at each node.
RESULTS
Maximum Likelihood Phylogeny of the Gesnerioideae—
The maximum likelihood analysis of the Gesnerioideae data
set resulted in a single most likely tree (−ln L = 17313.46185;
Fig. 1). ML bootstrap analysis resulted in similar support for
branches as previously found with parsimony analyses (Zim-
mer et al. 2002). This phylogenetic hypothesis resulted in all
seven tribes here included forming monophyletic groups
and, within the Gloxinieae, the two Central American clades
forming a grade leading to the South American Gloxinieae
clade (Fig. 1).
Maximum Likelihood and Bayesian Phylogenies and Tests
of Alternative Topologies in the Gloxinieae—The maximum
likelihood analysis of the Gloxinieae data set resulted in a
single most likely tree (−ln L = 8989.95793; Fig. 2). Bayesian
inference analyses resulted in plots of log-likelihood scores
and all other parameters reaching stationarity prior to gen-
eration 20,000,000 in all independent analyses (plots not
shown). The first 200,001 sample points were thus discarded
as burn-in, leaving 800,000 samples for construction of a 50%
majority rule consensus tree. Figure 2 illustrates the ML to-
pology with 50% posterior probability distributions from one
of the runs mapped onto the phylogeny (results from this run
are used in all further results and discussions). All five inde-
pendent analyses resulted in similar posterior probability
distributions and therefore we expect that convergence and
mixing is occurring (data not shown). Sixty-four percent of
nodes have a posterior probability ⱖ95%. The ML and
Bayesian topologies are congruent but with some small dif-
ferences either at unsupported nodes, or among closely re-
lated species. Both of these analyses of the combined molecu-
lar dataset present the two Central American clades as a
grade leading to the South American Gloxinieae clade (Fig.
2). However, when we test the alternative topology with the
Central American clades forced to form a sister clade to the
South American clade, we do not find a significant difference
in likelihood score, as measured by the SH test (P = 0.4221).
This is not unexpected as only a marginal majority of trees
from the Bayesian posterior probability distribution of trees
supported this branch of the highest likelihood tree (PP =
52%; Fig. 1).
Reconstruction of Biogeographic Patterns—The FPO bio-
geographic reconstruction of subfamily Gesnerioideae re-
sulted in two different topologies, depending on whether the
Achimenes/Solenophora and Eucodonia/Smithiantha/Moussonia/
Niphaea clades (referred to in the future as the “Central
American clades”) are sister to each other or form a grade
leading to the rest of the tribe. When the Central American
clades form a grade leading to the rest of the Gloxinieae, the
ancestral node of the Gloxinieae (and most of the rest of the
internal nodes of the phylogeny) is equivocal (Fig. 3A). When
these two clades are considered sister to one another, the
ancestor of the Gloxinieae tribe is considered to be of An-
dean/western South American origin (Fig. 3B). Similarly, the
ancestor of the Gesnerieae/Gloxinieae clade, the ancestor of
the Episcieae + Sinningieae + Sphaerorrhiza sarmentiana + Ges-
nerieae + Gloxinieae clade, and the ancestor of subfamily
Gesnerioideae are all reconstructed as being of Andean/
western South American origin (Fig. 3B). The ancestral nodes
of the Episcieae, Sinningieae, and Sphaerorrhiza sarmentiana
lineages are all equivocal, and can be reconstructed as An-
dean/western South American or southern Brazilian in ori-
gin.
The DIVA Gesnerioideae biogeographic reconstruction re-
sulted in 12 inferred dispersal/extinction events regardless of
topology where the maximum areas for each node was un-
constrained (Fig. 3A, B). When the Central American clades
form a grade leading to the rest of the Gloxinieae, the ances-
tor of the Gloxinieae is reconstructed as being of Central
American origin (Fig. 3A). However, where these clades are
sister to one another (Fig. 3B), the ancestor of the Gloxinieae
tribe is reconstructed as widespread through Central
America and Andean/western South America or restricted to
Central America. When the maximum areas at each node
were restricted to two, either 15 or 16 dispersal events were
necessary to reconstruct biogeographic patterns (Fig. 3A, B).
The reconstruction of the ancestral state of the Gloxinieae
was restricted to either a Central American origin or wide-
spread in Central America and Andean/western South
America if the Central American clades form a grade (Fig.
3A), but the reconstruction of this node when they are in-
ferred to be sister clades was a widespread Andean/western
South American and Central American distribution (Fig. 3B).
Biogeographic analysis of the Gloxinieae with FPO re-
sulted in two different reconstructions of biogeographic pat-
terns, both of six steps (Fig. 4A, B). The only areas of ambi-
guity in the reconstruction is where the Central American
clades form a grade is at the root (Fig. 4A), and, where the
Central American clades form a clade, the only ambiguous
reconstructions are at the ancestor of the Gloxinieae and the
root (Fig. 4B). The only differences between the two FPO
patterns are the reconstruction of the ancestral state for tribe
Gloxinieae and the number of alternative reconstructions at
the Gloxinieae ancestor and root nodes.
DIVA analysis of the Gloxinieae topologies resulted in two
possible patterns. DIVA pattern 1 resulted in an optimum
reconstruction of five dispersal events whether constrained
to two ancestral areas or unconstrained (Fig. 4A). DIVA pat-
tern 2 (Fig. 4B) resulted in an optimum reconstruction of four
TABLE 1. Constraint sets for r8s analyses. All times are given in mil-
lions of years ago (MYA). Node numbers refer to Fig. 5.
Constraint Node 56 Node 52 Node 32 Node 8
56/52 Max = 71 Max = 35 — —
Min=25
56/32 Max = 71 — Max = 35 —
Min=25
56/8 Max = 71 — — Max = 16
56/52/8 Max = 71 Max = 35 — Max = 16
56/32/8 Max = 71 — Max = 35 Max = 16
162 SYSTEMATIC BOTANY [Volume 33

FIG. 1. Maximum likelihood phylogeny of Gesnerioideae relationships (−lnL = 17313.46185). ML bootstrap values are noted above branches. Open
circles on branches refer to branches with a maximum parsimony bootstrap percentage ⱖ70% and closed circles on branches refer to branches with a
maximum parsimony bootstrap percentage ⱖ95% from previous analyses of the combined ITS/trnL–F/trnE–Tcombined dataset (Zimmer et al. 2002).
Shaded boxes outline the two Central American Gloxinieae clades. Bars to the right of clades denote tribes abbreviated as follows: B = Beslerieae, E =
Episcieae, Ge = Gesnerieae, Gl = Gloxinieae, N = Napeantheae, Si = Sinningieae, and Sp = Sphaerorrhizeae.
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 163

FIG. 2. Maximum likelihood phylogeny of Gloxinieae relationships (−lnL = 8989.95793) based on the combined ITS/trnL–Fdataset. Bayesian
inference posterior probability values are noted above branches. Open circles on branches refer to branches with a maximum parsimony bootstrap
percentage ⱖ70% and closed circles on branches refer to branches with a maximum parsimony bootstrap percentage ⱖ95% from previous analyses of
the combined ITS/trnL–F/ morphology combined dataset (Roalson et al. 2005a). Shaded boxes outline the two Central American Gloxinieae clades. Bars
to the right of clades denote tribes abbreviated as follows: Ge = Gesnerieae and Gl = Gloxinieae.
164 SYSTEMATIC BOTANY [Volume 33

dispersal events when unconstrained, or five dispersals
when constrained to two ancestral areas. The ancestral node
of the Gloxinieae is alternatively resolved as either having a
Central American distribution or widespread in Andean/
western South America and Central America (Fig. 4A, B).
Estimating Absolute Divergence Times of Gesnerioideae
Clades—The five constraint sets for estimating absolute di-
vergence times in the Gesnerioideae all support very similar
nodal ages (Table 2). Figure 5 diagrams the estimated nodal
ages and confidence intervals under one of the five constraint
sets (56/52/8). It should be noted, however, that the confi-
dence intervals under all of the constraint sets overlap, sug-
gesting that there is likely no significant difference among
node age estimates under different constraints (Table 2). If
the maximal age of the subfamily is less than 71 MYBP, as
suggested by Bremer et al. (2004), the constraint of nodes
associated with movements between South America, Central
America, and the Caribbean to the estimated timing of land
bridge or narrow land mass gaps seems to fit the evolution-
ary divergence of the genes sampled and their estimated rate
as inferred by the penalized likelihood smoothing methods.
DISCUSSION
ML and Bayesian Inference Analyses of Gloxinieae Rela-
tionships—ML and Bayesian inference analyses of the phy-
logenetic relationships within the Gloxinieae provide a rela-
tively well-resolved and well-supported phylogenetic hy-
pothesis (Fig. 2). These results are generally congruent with
previous analyses of these datasets using maximum parsi-
mony (Roalson et al. 2005a). This is illustrated by the maxi-
mum parsimony bootstrap percentages mapped onto the ML
phylogenetic hypothesis (Fig. 2). No branches previously
considered moderately or well supported by bootstrap per-
centages above 70% or 95% are not present in the ML tree,
and most branches with >70% bootstrap support have high
or moderate Bayesian posterior probability values (Fig. 2;
Roalson et al. 2005a). Additionally, several branches without
parsimony bootstrap support are strongly supported by the
Bayesian posterior probability values (Fig. 2). While there has
been some controversy regarding the interpretation of Bayes-
ian posterior probability values, we consider them to provide
the confidence interval around each branch. Given the con-
sistent posterior probability distributions postburnin from
different start locations in multiple independent Bayesian
analyses, we suggest that the posterior probability distribu-
tion of trees and parameters are adequately sampled and
reflect the true parameter distribution. Therefore, as we are
confident that our model choice was appropriate given the
datasets, we expect that the posterior probability scores re-
flect the 95% confidence interval for tree branches. It has been
previously noted that posterior probability values are some-
times able to provide high confidence on branches with low
bootstrap percentages when internodes are short due to an
apparent greater sensitivity of Bayesian PP to signal in the
datasets (Alfaro et al. 2003). As with the branches also sup-
ported by high bootstrap values, we consider these branches
with high posterior probability values, but low bootstrap
support, to be well supported (Fig. 2).
Reconstruction of Biogeographic Patterns—Biogeographic
patterns in the Gesneriaceae rarely have been explored, and,
traditionally, discussions of biogeographic origins and pat-
terns in the family have relied on association of areas of
highest diversity with an implied area of origin (Gentry 1982;
Kvist and Skog 1992, 1996; Skog and Kvist 2000). No explicit
biogeographic reconstructions have previously been made in
the New World Gesnerioideae, but this group has been
thought to have originated in northern South America due to
the high diversity of lineages of the subfamily in this region
and it has been suggested that lineages in Central America
and the Caribbean Islands have resulted from multiple dis-
persal events from South America. The results presented here
suggest several alternative reconstructions of the biogeo-
graphic patterns in the New World Gesnerioideae and the
Gloxinieae tribe.
In the FPO and DIVA analyses of the Gesnerioideae and
Gloxinieae, the reconstruction of the ancestral distribution
for the Gloxinieae is entirely dependent on whether the two
Central American clades form a grade leading to the rest of
the Gloxinieae or are sister clades (Figs. 3, 4). Unfortunately,
these nodes are not well supported in the ML and Bayesian
analyses (Figs. 1, 2) and the alternative hypothesis of a mono-
phyletic Central American clade cannot be excluded based
on tests of likelihood scores using the SH test. Given the low
support of one of these hypotheses over the other, it is best to
consider these two topologies equally likely.
While this lack of support of one hypothesis over the other
could be interpreted as uninformative, we believe an alter-
native perspective is more appropriate. Historically, biogeo-
graphic hypotheses have been formed entirely on the regions
of highest diversity argument. These analyses suggest two
equally plausible (given the data) competing biogeographic
hypotheses which can be tested with additional data and
sampling in the future: (1) the Gloxinieae tribe originated in
Central America; or (2) the ancestor of the Gloxinieae had a
widespread distribution, including both Central America
and Andean/western South America. Additional gene se-
quences (such as low copy number nuclear genes or more
variable chloroplast spacers) may help resolve weak nodes
and thus narrow down which hypothesis might be most
likely. It seems clear that the likelihood of the competing
hypotheses is dependent on the resolution of a few nodes
associated with the relationships among the two clades of
predominantly Central American genera and the large and
strongly supported predominantly Andean/western South
American clade of Gloxinieae genera. Other lineages in this
clade found in Central America clearly appear to be separate
dispersal events from South America as they are nested
within predominantly South American clades of species. It
should be noted that while some reconstructions suggest the
Gloxinieae originated in Central America, none of these
analyses suggest the ancestor was restricted to Andean/
western South America (Figs. 3, 4). This is notable in that
while confidence in some of the nodes of interest might be
low; all indications suggest that this lineage was in Central
America prior to the divergence of Gloxinieae lineages that
are currently extant.
Other recent phylogenetic studies of the Gloxinieae based
on a combination of these and other gene regions (Smith et al.
2004a, b) have resulted in phylogenetic hypotheses that differ
significantly from the phylogenetic results found here and
elsewhere (Zimmer et al. 2002; Roalson et al. 2005a). This
conflict has been previously found in relation to other Smith
et al. studies (particularly Smith 2000b) by Zimmer et al.
(2002), who noted that the exact same samples sequenced
differed by as much as 4% and had a large number of am-
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 165

FIG. 3. FPO and DIVA optimizations of biogeographic patterns in the Gesnerioideae. Colored circles at each node represent FPO reconstructions and
letters note DIVA reconstructions as follows: white&A=OldWorld; dark blue&B=Andean/western South America; light blue&C=Central
America; green&D=Caribbean; yellow&E=southern Brazil; and black&F=Guianas. Nodes without letter designations have the same state
reconstruction using DIVA and FPO. Letters in parentheses refer to DIVA reconstructions under the “maximum of two ancestral areas”constraint. Letter
combinations divided by a “/”reflect alternative DIVA reconstructions. Bars above clades denote tribes abbreviated as follows: B = Beslerieae, E =
166 SYSTEMATIC BOTANY [Volume 33

biguous base calls. While polymorphism has been found in
ITS sequences among many plant lineages, significant poly-
morphism within species and individuals has not been found
in Gesnerioideae (particularly Gesnerieae and Gloxinieae)
ITS sequences (unpubl. data). In assessing the different place-
ment of species in previous analyses, some taxa have been
clearly misplaced in these studies, notably the placement of
Sanango within the Gesnerieae tribe (Smith et al. 1997a; erro-
neous placement discussed by Weber 2004), placement of
Lembocarpus within the Gloxinieae tribe (Smith 2001; place-
ment clarified by Smith et al. 2004b; Roalson et al. 2005a; this
paper), and placement of Capanea grandiflora (Kohleria tigridia)
in the Episcieae tribe (Smith 2001; placement clarified by
Roalson et al. 2005a; this paper). As with these previous is-
sues, more recent Smith et al. (2004a, b) publications have
similarly conflicting phylogenies. Particularly, the placement
of Solenophora sister to Niphaea,Koellikeria (Gloxinia)erinoides
sister to Goyazia,Anodiscus (Gloxinia)xanthophyllus sister to
Seemannia (all in Smith et al. 2004a), and Niphaea sister to
Gloxinia,Diastema sister to Gesneria, and Phinaea sister to Glox-
inia (all in Smith et al. 2004b) are problematic as their place-
ment is very different from other studies (Zimmer et al. 2002;
Roalson et al. 2005a; this study), and also inconsistent among
the different Smith et al. publications. As an example, Dia-
stema racemiferum is placed sister to Gesneria in Smith et al.
2004a and sister to Heppiella in Smith et al. 2004b. The move-
ment between these branches requires crossing several
strongly supported branches as measured by bootstrap and
Bayesian posterior probability support values, both in the
Smith et al. papers as well as Roalson et al. 2005a and this
study, and moving between different tribes of the Gesneri-
oideae. However, many of the branches in the Smith et al.
(2004a, b) studies are not strongly supported, and therefore
some of these differences may be related to short branches as
well as possible conflicting signal in different genic regions.
Smith et al. (2004b) do note significant incongruence in their
datasets as measured by the partition homogeneity test.
Given the lower species sampling within the Gloxinieae in
these papers and the use of eight genic regions, it is unclear
why there would be less clade support and different relation-
ships from our study using two genic regions, both of which
are used in the Smith et al. (2004b) analyses. Our laboratory
techniques produce sequences with strong and consistent
phylogenetic signal. Until the cause of the discrepancies
shown by the other gene regions analyzed by Smith et al.
(2004a, b) is understood, we prefer to exclude this data from
our analysis.
Dating Divergences, Plate Tectonics, and Phylogenetic
Patterns—Optimally, we could test the two alternative hy-
potheses by comparing them to estimates of the ages of an-
cestral nodes and the timing of continental movements. Un-
fortunately, there is effectively no fossil record for the entire
family Gesneriaceae, let alone a much more recently diversi-
fying Gloxinieae tribe. This allows no point for absolute date
calibration in the near vicinity of the Gloxinieae tribe. Recent
studies of dating divergences have used multiple calibration
points outside of the lineage of interest to effectively calibrate
nodes with no known fossil record (Yoder and Yang 2004).
This kind of study is possible for the Gesneriaceae, but only
through analysis of a reasonable diversity within the entire
Gesneriaceae as well as sampling from throughout the Lami-
ales. Unfortunately, this dataset does not currently exist, but
it would be possible in the future to use such a dataset to test
the timing implications of our competing biogeographic hy-
potheses. Bremer et al. (2004) have recently applied similar
methods to place the stem age of the Gesneriaceae at approxi-
mately 71 MYBP, which is similar to the 65 MYBP estimate
made by Raven and Axelrod (1974). Most of the estimates of
the age of the Gesnerioideae did not approach this family
crown maximum, given other calibration points, and the es-
timated subfamily crown age of 57 MYBP (Table 2; Fig. 5)
seems reasonable if the major lineages of the family diverged
in the late Cretaceous to early Tertiary. Additionally, the age
estimates for divergence of many of the genera in the Epi-
scieae, Gloxinieae, and Sinningieae span the Miocene and
Oligocene while those for the Gesnerieae are predominantly
Pliocene and late Miocene (Fig. 5). These differences and
similarities in divergence of recognized genera need to be
further explored with regards to vegetation patterns in Cen-
tral and South America and the Caribbean during these time
periods.
Alternatively, the timing of continental movements and
other geological events can be used to calibrate the timing of
biogeographic events (Baldwin and Sanderson 1998; Renner
et al. 2000; Bossuyt and Milinkovitch 2001; Cooper et al. 2001;
Lavin et al. 2001; Richardson et al. 2001; Conti et al. 2002;
Morley and Dick 2003). Here we have used five different
constraint sets (Table 1) and these constraints all provide
similar estimates of node age across the phylogeny. Our ap-
plication of the hypothesis that the movement of gesneriad
lineages among South America, Central America, and the
Caribbean prior to ∼3 MYBP was mediated by the presence of
GAARLANDIA and the proximity of the southern Central
American land mass to the northwest South American mi-
crocontinent (as described by Iturralde-Vinent and MacPhee
1999) seems to fit well with our estimates of nodal ages.
While the use of geological events alone is somewhat limiting
for the inference of timing of dispersal, in the event of a
complete absence of a fossil record, there is little else to use to
infer these patterns. It should be noted that the maximal age
of the clade is derived from a fossil-calibrated study (Bremer
et al. 2004), and the various other calibration points do not
give significantly different ages, suggesting that the applica-
tion of these maximal ages or max-min ranges are not unduly
influencing (or applying a circular reasoning argument to)
the inferred node ages.
It should be noted that the application of the above time
calibrations to particular nodes/events requires several as-
sumptions about the mechanisms of dispersal of gesneriads.
In order to use the timing of continental movements, and
particularly the formation of the GAARLANDIA land bridge,
we have to assume that the formation of this bridge in some
way positively influenced the opportunities for dispersal of
gesneriads between these two land masses.
←
Episcieae, Ge = Gesnerieae, Gl = Gloxinieae, N = Napeantheae, Si = Sinningieae, and Sp = Sphaerorrhizeae. A. Character reconstruction with a Central
America grade of clades in the Gloxinieae. Twelve dispersals inferred in unconstrained analyses and 16 inferred dispersals when constrained to two
ancestral areas. B. Character reconstruction with a monophyletic Central America clade in the Gloxinieae. Twelve dispersals inferred in unconstrained
analyses and 15 inferred dispersals when constrained to two ancestral areas.
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 167

168 SYSTEMATIC BOTANY [Volume 33

TABLE 2. Reconstruction of nodal ages using penalized likelihood. Nodes labeled1–56 refer to nodes from Fig. 5. Details of the constraints are
described in the text and Table 1. All times are given in millions of years ago (MYA). Numbers in parentheses are the minimum and maximum values
for the 95% confidence interval, respectively.
Node Constraint 56/52 Constraint 56/32 Constraint 56/8 Constraint 56/52/8 Constraint 56/32/8
1 7 (4.76, 9.04) 8.9 (5.83, 11.67) 6.9 (3.69, 9.69) 6.9 (3.71, 9.27) 6.9 (3.52, 9.8)
2 8.5 (6.36, 10.64) 11 (7.75, 13.83) 8.4 (4.91, 11.55) 8.4 (4.92, 11.08) 8.4 (4.64, 11.76)
3 11 (8.21, 13.41) 14 (10.23, 17.19) 10 (6.57, 14.33) 10 (6.49, 13.85) 10 (6.18, 14.62)
4 6.8 (4.91, 8.71) 8.6 (6.13, 11.05) 6.7 (3.87, 9.31) 6.7 (3.86, 8.98) 6.7 (3.8, 9.36)
5 9.8 (7.74, 12.06) 12 (9.77, 15.21) 9.6 (6.07, 12.99) 9.6 (6.03, 12.59) 9.6 (5.89, 13.13)
6 15 (12.85, 17.25) 19 (16.02, 21.94) 15 (10.04, 18.72) 15 (9.82, 18.38) 15 (9.65, 19.05)
7 14 (12.1, 16.42) 18 (15.09, 20.89) 14 (9.45, 17.85) 14 (9.24, 17.52) 14 (8.92, 18.28)
8 16 (14.2, 18.56) 21 (17.7, 23.7) 16 (11.09, 20.13) 16 (10.79, 19.87) 16 (11.2, 20.04)
9 1.8 (0.68, 3.04) 2.3 (0.9, 3.78) 1.8 (0.55, 3.07) 1.8 (0.53, 2.97) 1.8 (0.53, 3.13)
10 6.5 (4.74, 8.3) 8.2 (5.97, 10.49) 6.4 (3.79, 8.91) 6.4 (3.74, 8.58) 6.4 (3.54, 9.1)
11 8.9 (6.89, 10.93) 11 (8.51, 14.03) 8.8 (5.46, 11.9) 8.7 (5.41, 11.45) 8.8 (5.43, 11.91)
12 16 (13.81, 18.53) 21 (17.42, 23.54) 16 (12.76, 18.96) 16 (12.47, 18.47) 16 (12.09, 19.53)
13 17 (14.48, 18.84) 21 (17.86, 24.26) 17 (10.28, 21.6) 17 (10.07, 20.99) 17 (9.89, 21.89)
14 18 (15.49, 19.97) 23 (19.52, 25.48) 17 (13.57, 20.97) 17 (13.22, 20.5) 17 (13.57, 20.97)
15 10 (7.89, 12.57) 13 (9.67, 16.19) 10 (6.35, 13.55) 9.9 (6.24, 13.08) 10 (5.97, 13.85)
16 13 (11.05, 15.73) 17 (13.72, 20.12) 13 (8.77, 17.25) 13 (8.66, 16.62) 13 (8.23, 17.67)
17 18 (16.06, 20.78) 23 (19.89, 26.65) 18 (12.67, 23.03) 18 (12.41, 22.37) 18 (12.74, 22.98)
18 0.72 (0, 1.73) 0.94 (0, 2.19) 0.72 (0, 1.75) 0.72 (0, 1.67) 0.72 (0, 1.75)
19 2.5 (0.93, 3.97) 3.3 (1.23, 5.03) 2.5 (0.61, 4.17) 2.5 (0.6, 4) 2.5 (0.65, 4.17)
20 4.1 (2.82, 5.34) 5.2 (3.64, 6.8) 4.1 (2.44, 5.6) 4 (2.4, 5.36) 4.1 (2.5, 5.58)
21 3.9 (2.17, 5.33) 5 (2.73, 6.85) 3.9 (1.59, 5.59) 3.9 (1.59, 5.39) 3.9 (1.45, 5.69)
22 5.4 (3.86, 6.66) 7 (4.94, 8.62) 5.4 (3.3, 7.18) 5.4 (3.22, 6.82) 5.4 (3.3, 7.18)
23 6.1 (4.3, 7.42) 7.8 (5.54, 9.38) 6.1 (3.81, 7.77) 6 (3.82, 7.38) 6.1 (3.94, 7.7)
24 3.4 (1.94, 4.86) 4.4 (2.41, 6.29) 3.4 (1.66, 5.02) 3.4 (1.62, 4.82) 3.4 (1.54, 5.1)
25 7.2 (4.95, 8.75) 9.3 (6.34, 11.14) 7.2 (4.66, 9.06) 7.1 (4.62, 8.62) 7.2 (4.66, 9.06)
26 0.35 (0, 0.79) 0.45 (0, 1.02) 0.35 (0, 0.96) 0.35 (0, 0.93) 0.35 (0, 0.96)
27 5.7 (4.14, 7.54) 7.4 (5.33, 9.57) 5.7 (3.4, 8.16) 5.7 (3.38, 7.74) 5.7 (3.2, 8.32)
28 2.9 (1.56, 4.28) 3.7 (2.1, 5.34) 2.9 (1.25, 4.53) 2.8 (1.26, 4.3) 2.9 (1.16, 4.6)
29 4.2 (2.67, 5.75) 5.4 (3.32, 7.48) 4.2 (2.15, 6.19) 4.1 (2.24, 5.8) 4.2 (1.99, 6.31)
30 6.3 (4.64, 8.16) 8.1 (5.9, 10.46) 6.3 (3.9, 8.78) 6.3 (3.86, 8.34) 6.3 (3.69, 8.93)
31 8.5 (6.24, 10.68) 11 (7.84, 13.76) 8.5 (5.22, 11.54) 8.4 (5.19, 10.95) 8.5 (5.43, 11.43)
32 26 (22.47, 29.47) 33 (28.45, 36.97) 26 (17.74, 33.38) 26 (17.45, 31.97) 26 (17.76, 33.36)
33 9.3 (7.28, 12.12) 12 (8.96, 15.4) 9.4 (5.71, 13.51) 9.3 (5.71, 12.79) 9.4 (5.46, 13.7)
34 9.9 (7.75, 11.95) 13 (9.55, 15.35) 9.9 (5.61, 13.69) 9.9 (5.57, 13.09) 9.9 (5.61, 13.69)
35 11 (8.96, 12.92) 13 (10.99, 16.51) 11 (6.94, 14.74) 11 (6.92, 13.96) 11 (6.96, 14.72)
36 13 (10.89, 15.73) 16 (13.35, 20.07) 13 (8.4, 17.96) 13 (8.38, 17.02) 13 (8.4, 17.96)
37 15 (12.24, 18.16) 19 (15.16, 23.08) 15 (9.48, 20.72) 15 (9.49, 19.57) 15 (9.46, 20.74)
38 18 (14.34, 22.18) 23 (17.85, 28.17) 18 (10.99, 25.27) 18 (11.05, 23.81) 18 (10.44, 25.68)
39 21 (18.18, 24.3) 27 (22.45, 31.01) 21 (13.69, 28.37) 21 (13.65, 26.81) 21 (13.33, 28.65)
40 8.3 (5.68, 11.76) 11 (7.14, 14.86) 8.3 (4.65, 12.69) 8.3 (4.63, 12.07) 8.3 (4.45, 12.85)
41 23 (20.08, 26.64) 30 (24.5, 34.42) 23 (15.2, 31.24) 23 (15.18, 29.42) 23 (14.75, 31.59)
42 12 (9.44, 14.92) 15 (11.59, 18.91) 12 (7.35, 16.87) 12 (7.38, 15.94) 12 (6.97, 17.17)
43 19 (16.12, 22.44) 24 (19.82, 28.5) 19 (12.3, 26.02) 19 (12.37, 24.49) 19 (11.63, 26.55)
44 25 (22.78, 28.46) 32 (27.58, 36.82) 26 (16.71, 34.39) 25 (16.93, 32.13) 26 (17.08, 34.12)
45 28 (25.39, 30.31) 35 (30.54, 39.58) 28 (18.55, 36.99) 28 (18.75, 34.59) 28 (18.9, 36.74)
46 14 (11.06, 15.82) 17 (13.63, 20.39) 14 (8.31, 18.47) 14 (8.42, 17.3) 14 (8.31, 18.47)
47 15 (12.02, 17.06) 18 (14.82, 21.82) 15 (9.26, 19.58) 14 (9.34, 18.38) 15 (8.63, 20.07)
48 17 (14.38, 19.02) 21 (17.56, 24.48) 17 (10.79, 22.35) 17 (10.8, 21.04) 17 (10.05, 22.93)
49 18 (15.67, 20.55) 23 (19.14, 26.46) 18 (11.99, 24.03) 18 (11.98, 22.66) 18 (12.49, 23.81)
50 32 (28.7, 33.58) 40 (33.86, 44.42) 32 (18.21, 42.93) 32 (18.33, 40.25) 32 (19.18, 42.34)
51 32 (30.07, 34.03) 41 (35.66, 45.26) 32 (21.94, 42.1) 32 (22.01, 39.45) 32 (22.45, 42.33)
52 35 (35, 35) 45 (39.64, 48.96) 35 (24.35, 45.79) 35 (24.46, 42.74) 35 (25.09, 45.73)
53 15 (12.07, 18.47) 19 (15.08, 23.08) 15 (10.09, 20.41) 15 (10.05, 19.29) 15 (9.5, 20.86)
54 17 (13.24, 20.88) 21 (16.62, 25.98) 17 (10.86, 23.1) 17 (10.82, 21.86) 17 (10.18, 23.62)
55 52 (43.68, 60.44) 65 (57.22, 71.62) 52 (33.63, 69.51) 52 (33.65, 65.45) 52 (31.09, 71.49)
56 56.89 (51.04, 63.88) 71 (70.02, 71.78) 57.23 (39.1, 74.86) 56.64 (38.88, 70.76) 57.23 (40.23, 74.15)
←
FIG. 4. FPO and DIVA optimizations of biogeographic patterns in the Gloxinieae. Colored circles at each node represent FPO reconstructions and
letters note DIVA reconstructions as follows: white&A=Andean/western South America; blue&B=Central America; green&C=Caribbean; and
black&D=southern Brazil. Nodes without letter designations have the same state reconstruction using DIVA and FPO. Letters in parentheses refer
to DIVA reconstructions under the “maximum of two ancestral areas”constraint. Letter combinations divided by a “/”reflect alternative DIVA
reconstructions. Bars above clades denote tribes abbreviated as follows: Ge = Gesnerieae and Gl = Gloxinieae. A. Character reconstruction with a Central
America grade of clades in the Gloxinieae. Five dispersals inferred in unconstrained analyses and five inferred dispersals when constrained to two
ancestral areas. B. Character reconstruction with a monophyletic Central America clade in the Gloxinieae. Four dispersals were inferred in the
unconstrained analyses, and five when constrained to two ancestral areas.
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 169

Nothing definitive is known about mechanisms of dis-
persal in New World gesneriads and circumstantial infer-
ences of possible dispersal vectors suggest that there are sev-
eral potential vectors including wind, animals (both through
consumption and passive dispersal), and water. It is not clear
how each of these dispersal vectors would be influenced by
the formation of a land bridge. Endozoochory seems unlikely
for many Gesneriaceae species given the small seed size in
FIG. 5. Chronogram reflecting divergence times in millions of years ago (MYA) using the 56/52/8 constraints. Circled numbers refer to node divergence
calculations listed in Table 2. Shaded bars reflect the bootstrap 95% confidence interval described in Table 2. Bars to the right of clades denote tribes
abbreviated as follows: B = Beslerieae, E = Episcieae, Ge = Gesnerieae, Gl = Gloxinieae, N = Napeantheae, Si = Sinningieae, and Sp = Sphaerorrhizeae.
170 SYSTEMATIC BOTANY [Volume 33

the family and the lack of a hard seed coat. If shorter distance
mechanisms such as splash cups and ant dispersal were
dominant dispersal types of ancestral lineages, we might ex-
pect that dispersal between the Caribbean Islands and South
America would be positively influenced by GAARLANDIA.
Alternatively, if wind dispersal or long distance dispersal
passively or internally by birds were the more dominant dis-
persal mechanisms, we might expect a lack of correlation of
long distance dispersal among Caribbean islands, Central
America, and South America and the formation of GAAR-
LANDIA or other island stepping stone formations.
So what do we know about the distribution of dispersal
mechanisms in the Gesneriaceae? Several authors have pro-
vided anecdotal hypotheses of dispersal mechanisms in dif-
ferent lineages in the family. In the Old World, Burtt (1970,
1976), Carlquist (1970), Sakai et al. (1995), Denduangboripant
and Cronk (2000), Denduangboripant et al. (2001), Mendum
et al. (2001), Kiehn (2001), Price and Wagner (2004), and
Cronk et al. (2005) have discussed dispersal mechanisms pri-
marily in regard to interisland dispersal of Aeschynanthus and
Cyrtandra among southeast Asian and Pacific island land-
masses. They particularly note that the coma of filiform ap-
pendages in some Aeschynanthus and the white fleshy fruits
of the Pacific island lineages of Cyrtandra might play roles in
their success in dispersal through wind and putative bird
endozoochory, respectively. Burtt (1970, 1976) also noted the
likely dispersal of understory herbs through water dispersal
mechanisms, such as splash cups and adherence of the tiny
seeds to animals.
In the New World, evidence for dispersal is largely asso-
ciated with the genera Columnea,Gasteranthus,Gesneria, and
Kohleria (Stearn 1969; Skog and Kvist 2000; Skog 1976; and
Kvist and Skog 1992; respectively). Fruits in Jamaican Colum-
nea species are presumably dispersed by bird consumption
(Stearn 1969), although others have suggested these seeds
might as likely be dispersed by ants (attributed to Morley by
Stearn 1969). Gasteranthus generally have fleshy capsules that
split to reveal a mass of seeds. The dispersal mechanism for
these seeds is unclear, but rain wash in the very wet habitats
where Gasteranthus is found and accidental sticking of the
tiny seeds to passing animals are the most likely contributors
to seed dispersal (Skog and Kvist 2000). Given the much
more widespread distribution of Gasteranthus pansamalanus
compared to most other Gasteranthus species, Skog and Kvist
(2000) suggested these seeds might be dispersed by bird con-
sumption of the berry-like capsules.
Skog (1976) outlined three likely mechanisms of dispersal
for Gesneria in his revision of the Gesnerieae. He considered
that transport of the small seeds in dry capsules would only
be carried incidentally by large animals, but that inverte-
brates and particularly ants might be important dispersers.
Plants growing near or in streams might likely have seeds
carried downstream by water in addition to the splash cup
dispersal found in some species. Finally, dispersal by wind
seems likely to be a major mechanism of dispersal in Gesneria.
He further notes that dispersal by animals and water seem to
be primarily short distance dispersal mechanisms with dis-
persal by wind the only opportunity for long distance dis-
persal of Gesneria seeds among the different Caribbean is-
lands. Kohleria appears to include species with both wind and
animal dispersal of seeds (Kvist and Skog 1992). The puta-
tively wind-dispersed species are generally quite widespread
and have small dust-like seeds in dehiscent dry capsules.
Other species have seeds presented in glutinous seed masses
that are presumably dispersed by animals, including insects
that may prey on seed masses and become covered in the
process.
An additional process possibly involved in dispersal of
gesneriads in the New World is hurricanes. Given the pre-
vailing patterns of hurricane movement –from the southeast
to the northwest –this seems to be a potential contributor to
plant movement. It could be that we do not see adaptation for
long distance seed dispersal in the gesneriads for a reason,
that is, selection has been for shorter distance movement,
which is the more common dispersal pattern for population
distributions, and this is why most species of New World
Gesneriaceae have narrow range distributions. On the rare
occasions when hurricanes are strong enough and follow the
necessary trajectory, however, species might be moved with
little adaptation involved, as hurricane-force winds might
carry not only seeds, but fruits or large plant fragments as
well. So if hurricane-mediated dispersal has been involved in
moving gesneriads in the past, why do we not see more of the
northeastern South American gesneriad genera in the Carib-
bean? One potential explanation for this is that while hurri-
cane-mediated dispersal is possible, it is an extremely rare
event and/or the current large Caribbean land masses are too
far for this to be an effective mechanism. In this case, the
presence of the GAARLANDIA bridge might have allowed
for more effective hurricane-mediated dispersal.
If GAARLANDIA and the Central American land masses
did positively influence major dispersal events, it might be
expected that some of the midrange dispersal mechanisms
including wind (hurricane or otherwise) and ectozoochory
might have been important in these dispersal events. Given
the rarity of widely distributed species in the Gesnerioideae,
it seems more likely that shorter dispersal distances and the
presence of land bridges played a role in these movements.
The examples given here of potential dispersal mechanisms
and fruit/seed types covers most of the major variants we see
in the Gesnerioideae. Which types of fruit/seed types were
most likely associated with the branches where we think ma-
jor dispersal events were occurring needs further explora-
tion, but given that the Gesnerieae and Gloxinieae tribes are
dominated by dry capsules, it seems likely that our infer-
ences of dispersal mechanisms are accurate.
In conclusion, the phylogenetic and biogeographic hypoth-
eses presented here further support a reasonably resolved
phylogeny of Gesneriaceae tribe Gloxinieae and suggest that
the ancestor of the Gloxinieae tribe was either restricted to
Central America or broadly distributed through Central
America and western/Andean South America. Further, it ap-
pears likely that the Greater Antilles/Aves Ridge landbridge
played a role in the movement of the Gesnerieae and Glox-
inieae tribes from South America to Central America and the
Caribbean, with later back dispersals to South America in the
Gloxinieae. While sampling within tribe Episceae was rela-
tively low in these analyses, Fig. 5 suggests this tribe may
have diversified earlier (or had less extinction) than the
Gloxinieae or Gesnerieae lineages, and this should be further
explored. The relatively young age of the extant Gesnerieae
should also be further explored, as this may be associated
with greater extinction rates in a lineage restricted to islands.
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 171

ACKNOWLEDGMENTS. Special thanks to J. K. Boggan for helpful discus-
sions on previous versions of the manuscript and two reviewers for their
helpful comments.
LITERATURE CITED
Alfaro, M. E., S. Zoller, and F. Lutzoni. 2003. Bayes or bootstrap? A
simulation study comparing the performance of Bayesian Markov
Chain Monte Carlo sampling and bootstrapping in assessing phylo-
genetic confidence. Molecular Biology and Evolution 20: 255–266.
Atkins, H., J. Preston, and Q. C. B. Cronk. 2001. A molecular test of
Huxley’s line: Cyrtandra (Gesneriaceae) in Borneo and the Philip-
pines. Biological Journal of the Linnean Society 72: 143–159.
Baldwin, B. G. and M. J. Sanderson. 1998. Age and rate of diversification
of the Hawaiian silversword alliance (Compositae). Proceedings of the
National Academy of Sciences USA 95: 9402–9406.
Benton, M. J. and F. J. Ayala. 2003. Dating the tree of life. Science 300:
1698–1700.
Bossuyt, F. and M. C. Milinkovitch. 2001. Amphibians are indicators of
early Tertiary “out-of-India”dispersal of vertebrates. Science 292:
93–95.
Bremer, K., E. M. Friis, and B. Bremer. 2004. Molecular phylogenetic
dating of Asterid flowering plants shows early Cretaceous diversi-
fication. Systematic Biology 53: 496–505.
Bromham, L. and D. Penny. 2003. The modern molecular clock. Nature
Reviews. Genetics 4: 216–224.
Burnham, R. J. and A. Graham. 1999. The history of neotropical vegeta-
tion: new developments and status. Annals of the Missouri Botanical
Garden 86: 546–589.
Burtt, B. L. 1970. Studies in the Gesneriaceae of the Old World XXXIII.
Some species of Cyrtandra, chiefly Bornean. Notes from the Royal Bo-
tanical Garden Edinburgh 30: 23–42.
Burtt, B. L. 1976. Notes on rain-forest herbs. The Gardens’Bulletin, Sin-
gapore 29: 73–80.
Burtt, B. L. 1998. Climatic accommodation and phytogeography of the
Gesneriaceae of the Old World. Pp. 1–27 in Diversity and taxonomy of
tropical flowering plants, eds. P. Mathew and M. Sivadasan. Kerala,
India: Mentor Books.
Carlquist, S. 1970. Island biology. New York: Columbia University Press.
Coney, P. J. 1982. Plate tectonic constraints on the biogeography of
Middle America and the Caribbean region. Annals of the Missouri
Botanical Garden 69: 432–443.
Conti, E., T. Eriksson, J. Schönenberger, K. J. Sytsma, and D. A. Baum.
2002. Early Tertiary out-of-India dispersal of Crypteroniaceae: evi-
dence from phylogeny and molecular dating. Evolution 56: 1931–
1942.
Cooper, A., C. Lalueza-Fox, S. Anderson, A. Rambaut, J. Austin, and R.
Ward. 2001. Complete mitochondrial genome sequences of two ex-
tinct moas clarify ratite evolution. Nature 409: 704–707.
Cronk, Q. C. B., M. Kiehn, W. L. Wagner, and J. F. Smith. 2005. Evolution
of Cyrtandra (Gesneriaceae) in the Pacific Ocean: the origin of a su-
pertramp clade. American Journal of Botany 92: 1017–1024.
Denduangboripant, J. and Q. C. B. Cronk. 2000. High intraindividual
variation in internal transcribed spacer sequences in Aeschynanthus
(Gesneriaceae): implications for phylogenetics. Proceedings of the
Royal Society of London. Series B. Biological Sciences 267: 1407–1415.
Denduangboripant, J., M. Mendum, and Q. C. B. Cronk. 2001. Evolution
in Aeschynanthus (Gesneriaceae) inferred from ITS sequences. Plant
Systematics and Evolution 228: 181–197.
Felsenstein, J. 2004. PHYLIP: Phylogeny Inference Package. Version 3.6.
Seattle: Department of Genome Sciences and Department of Biology,
University of Washington.
Gentry, A. H. 1982. Neotropical floristic diversity: phytogeographical
connections between Central and South America, Pleistocene cli-
matic fluctuations, or an accident of the Andean orogeny? Annals of
the Missouri Botanical Garden 69: 557–593.
Huelsenbeck, J. P. and F. Ronquist. 2001. MrBayes: Bayesian inference of
phylogeny. Bioinformatics (Oxford, England) 17: 754–755.
Iturralde-Vinent, M. A. and R. D. E. MacPhee. 1999. Paleogeography of
the Caribbean region: implications for Cenozoic biogeography. Bul-
letin of the American Museum of Natural History 238: 1–95.
Kiehn, M. 2001. South Pacific and Hawaiian Cyrtandra: molecular and
micromorphological studies. Malayan Nature Journal 55: 21–27.
Knowlton, N., L. A. Weigt, L. A. Solórzano, D. K. Mills, and E. Berming-
ham. 1993. Divergence in proteins, mitochondrial DNA, and repro-
ductive compatibility across the Isthmus of Panama. Science 260:
1629–1632.
Kvist, L. P. and L. E. Skog. 1992. Revision of Kohleria (Gesneriaceae).
Smithsonian Contributions to Botany 79: 1–83.
Kvist, L. P. and L. E. Skog. 1996. Revision of Pearcea (Gesneriaceae).
Smithsonian Contributions to Botany 84: 1–47.
Lavin, M., M. F. Wojciechowski, A. Richman, J. Rotella, M. J. Sanderson,
and A. B. Matos. 2001. Identifying Tertiary radiations of Fabaceae in
the Greater Antilles: alternatives to cladistic vicariance analysis. In-
ternational Journal of Plant Sciences 162(Suppl.): S53–S76.
Lewis, P. O. 2001. Phylogenetic systematics turns over a new leaf. Trends
in Ecology & Evolution 16: 30–37.
Maddison, W. P. and D. R. Maddison. 2006. Mesquite: a modular system
for evolutionary analysis. Version 1.12 http://mesquiteproject.org.
Magallón, S. and M. J. Sanderson. 2001. Absolute diversification rates in
angiosperm clades. Evolution; International Journal of Organic Evolu-
tion 55: 1762–1780.
McLoughlin, S. 2001. The breakup history of Gondwana and its impact on
pre-Cenozoic floristic provincialism. Australian Journal of Botany 49:
271–300.
Mendum, M., P. Lassnig, A. Weber, and F. Christie. 2001. Testa and seed
appendage morphology in Aeschynanthus (Gesneriaceae): phytogeo-
graphical patterns and taxonomic implications. Botanical Journal of
the Linnean Society 135: 195–213.
Minin, V., Z. Abdo, P. Joyce, and J. Sullivan. 2003. Performance-based
selection of likelihood models for phylogeny estimation. Systematic
Biology 52: 674–683.
Möller, M. and Q. C. B. Cronk. 1997. Phylogeny and disjunct distribution:
evolution of Saintpaulia. Proceedings of the Royal Society of London.
Series B. Biological Sciences 264: 1827–1836.
Möller, M. and Q. C. B. Cronk. 2001. Phylogenetic studies in Streptocarpus
(Gesneriaceae): reconstruction of biogeographic history and distri-
bution patterns. Systematics and Geography of Plants 71: 545–555.
Morley, R. J. and C. W. Dick. 2003. Missing fossils, molecular clocks, and
the origin of the Melastomataceae. American Journal of Botany 90:
1638–1644.
Muller, J. 1981. Fossil pollen records of extant angiosperms. Botanical
Review 47: 1–142.
Olmstead, R. G., K.-J. Kim, R. K. Jansen, and S. J. Wagstaff. 2000. The
phylogeny of the Asteridae sensu lato based on chloroplast ndhF
gene sequences. Molecular Phylogenetics and Evolution 16: 96–112.
Oxelman, B., M. Backlund, and B. Bremer. 1999. Relationships of the
Buddlejaceae s.l. investigated using parsimony jackknife and branch
support analysis of chloroplast ndhF and rbcL sequence data. Sys-
tematic Botany 24: 164–182.
Perret, M., A. Chautems, R. Spichiger, G. Kite, and V. Savolainen. 2003.
Systematics and evolution of tribe Sinningieae (Gesneriaceae): evi-
dence from phylogenetic analyses of six plastid DNA regions and
nuclear ncpGS. American Journal of Botany 90: 445–460.
Price, J. P. and W. L. Wagner. 2004. Speciation in Hawaiian angiosperm
lineages: cause, consequence, and mode. Evolution 58: 2185–2200.
Ramírez Roa, M. A. 1987. Revision de Achimenes (Gesneriaceae). Thesis,
Universidad Nacional Autonoma de Mexico, D. F.
Raven, P. H. and D. I. Axelrod. 1974. Angiosperm biogeography and past
continental movements. Annals of the Missouri Botanical Garden 61:
539–673.
Renner, S. S., B. D. Foreman, and D. Murray. 2000. Timing transantarctic
disjuctions in the Atherospermataceae (Laurales): evidence from
coding and noncoding chloroplast sequences. Systematic Biology 49:
579–591.
Richardson, J. E., F. M. Weitz, M. F. Fay, Q. C. B. Cronk, H. P. Linder, G.
Reeves, and M. W. Chase. 2001. Rapid and ancient origin of species
richness in the Cape flora of South Africa. Nature 412: 181–183.
Roalson, E. H., J. K. Boggan, L. E. Skog, and E. A. Zimmer. 2005a. Un-
tangling the Gloxinieae (Gesneriaceae). I. Phylogenetic patterns and
generic boundaries inferred from nuclear, chloroplast, and morpho-
logical cladistic data sets. Taxon 54: 389–410.
Roalson, E. H., J. K. Boggan, and L. E. Skog. 2005b. Reorganization of
tribal and generic boundaries in the Gloxinieae (Gesneriaceae: Ges-
nerioideae) and the description of a new tribe in the Gesnerioideae,
Sphaerorrhizeae. Selbyana 25: 225–238.
Roalson, E. H., L. E. Skog, and E. A. Zimmer. 2003. Phylogenetic rela-
tionships and the diversification of floral form in Achimenes (Ges-
neriaceae). Systematic Botany 28: 593–608.
Rodríguez, F., J. F. Oliver, A. Marín, and J. R. Medina. 1990. The general
172 SYSTEMATIC BOTANY [Volume 33

stochastic model of nucletide substitutions. Journal of Theoretical Bi-
ology 142: 485–501.
Ronquist, F. 1996. DIVA version 1.1. Computer program and manual
available by anonymous FTP from Uppsala University (ftp.uu.se or
ftp.systbot.uu.se); http://www.ebc.uu.se/systzoo/research/diva/
diva.html
Ronquist, F. 1997. Dispersal-vicariance analysis: a new approach to the
quantification of historical biogeography. Systematic Biology 46: 195–
203.
Sakai, A. K., W. L. Wagner, D. M. Ferguson, and D. R. Herbst. 1995.
Biogeographical and ecological correlates of dioecy in the Hawaiian
flora. Ecology 76: 2530–2543.
Sanderson, M. J. 2002. Estimating absolute rates of molecular evolution
and divergence times: A penalized likelihood approach. Molecular
Biology and Evolution 19: 101–109.
Sanderson, M. J. 2004. r8s 1.70. Computer program and documentation
available from http://phylo.ucdavis.edu/r8s/r8s.html.
Sanderson, M. J. and J. A. Doyle. 2001. Sources of error and confidence
intervals in estimating the age of angiosperms from rbcL and 18S
rDNA data. American Journal of Botany 88: 1499–1516.
Shimodaira, H. and M. Hasegawa. 1999. Multiple comparisons of log-
likelihoods with applications to phylogenetic inference. Molecular
Biology and Evolution 16: 1114–1116.
Skog, L. E. 1976. A study of the tribe Gesnerieae, with a revision of
Gesneria (Gesneriaceae: Gesnerioideae). Smithsonian Contributions to
Botany 29: 1–182.
Skog, L. E. and L. P. Kvist. 2000. Revision of Gasteranthus (Gesneriaceae).
Systematic Botany Monographs 59: 1–118.
Smith, J. F. 1996. Tribal relationships within Gesneriaceae: a cladistic
analysis of morphological data. Systematic Botany 21: 497–513.
Smith, J. F. 2000a. A phylogenetic analysis of tribes Beslerieae and
Napeantheae (Gesneriaceae) and evolution of fruit types: parsimony
and maximum likelihood analyses of ndhF sequences. Systematic
Botany 25: 72–81.
Smith, J. F. 2000b. Phylogenetic resolution within the tribe Episcieae (Ges-
neriaceae): congruence of ITS and ndhF sequences from parsimony
and maximum-likelihood analyses. American Journal of Botany 87:
883–897.
Smith, J. F. 2000c. Phylogenetic signal common to three data sets: com-
bining data which initially appears heterogeneous. Plant Systematics
and Evolution 221: 179–198.
Smith, J. F. 2001. The phylogenetic relationships of Lembocarpus and Goy-a
zia (Gesneriaceae) based on ndhF sequences. Annals of the Missouri
Botanical Garden 88: 135–143.
Smith, J. F. and S. Atkinson. 1998. Phylogenetic analysis of the tribes
Gloxinieae and Gesnerieae (Gesneriaceae): data from ndhF se-
quences. Selbyana 19: 122–131.
Smith, J. F. and C. L. Carroll. 1997. A cladistic analysis of the tribe Epi-
scieae (Gesneriaceae) based on ndhF sequences: origin of morpho-
logical characters. Systematic Botany 22: 713–724.
Smith, J. F., K. D. Brown, C. L. Carroll, and D. S. Denton. 1997a. Familial
placement of Cyrtandromoea, Titanotrichum, and Sanango: three prob-
lematic genera of the Lamiales. Taxon 46: 65–74.
Smith, J. F., S. B. Draper, L. C. Hileman, and D. A. Baum. 2004a. A
phylogenetic analysis within tribes Gloxinieae and Gesnerieae (Ges-
nerioideae: Gesneriaceae). Systematic Botany 29: 947–958.
Smith, J. F., L. C. Hileman, M. P. Powell, and D. A. Baum. 2004b. Evolu-
tion of GCYC, a Gesneriaceae homolog of CYCLOIDEA, within Ges-
nerioideae (Gesneriaceae). Molecular Phylogenetics and Evolution 31:
765–779.
Smith, J. F., J. C. Wolfram, K. D. Brown, C. L. Carroll, and D. S. Denton.
1997b. Tribal relationships in the Gesneriaceae: evidence from DNA
sequences of the chloroplast gene ndhF. Annals of the Missouri Botani-
cal Garden 84: 50–66.
Stearn, W. T. 1969. The Jamaican species of Columnea and Alloplectus
(Gesneriaceae). Bulletin of the British Museum (Natural History), Botany
4: 179–236, pl. 14–21.
Stehli, F. G. and S. D. Webb (eds.). 1985. The great American biotic exchange.
New York: Plenum Press.
Swofford, D. 2001. PAUP*: Phylogenetic analysis using parsimony, ver-
sion 4.0b10. Sunderland: Sinauer Associates.
Tamura, K. and M. Nei. 1993. Estimation of the number of nucleotide
substitutions in the control region of mitochondrial DNA in humans
and chimpanzees. Molecular Biology and Evolution 10: 512–526.
Wang, C. N., M. Möller, and Q. C. B. Cronk. 2004. Phylogenetic position
of Titanotrichum oldhamii (Gesneriaceae) inferred from four different
gene regions. Systematic Botany 29: 407–418.
Weber, A. 2004. Gesneriaceae. Pp. 63–158 in The Families and Genera of
Vascular Plants. Vol. 7. Dicotyledons. Lamiales (except Acanthaceae incl.
Avicennniaceae), ed. K. Kubitzki. Berlin: Springer.
Weigend, M. and H. Förther. 2002. A revision of the Central American
genus Solenophora (Gesneriaceae). Harvard Papers in Botany 7: 37–78.
Wiehler, H. 1976. A report on the classification of Achimenes, Eucodonia,
Gloxinia, Goyazia, and Anetanthus (Gesneriaceae). Selbyana 1: 374–404.
Wiehler, H. 1983. A synopsis of the neotropical Gesneriaceae. Selbyana 6:
1–219.
Wikström, N., V. Savolainen, and M. W. Chase. 2001. Evolution in the
angiosperms: calibrating the family tree. Proceedings of the Royal So-
ciety of London. Series B. Biological Sciences 268: 2211–2220.
Yoder, A. D. and Z. Yang. 2004. Divergence dates for Malagasy lemurs
estimated from multiple gene loci: geological and evolutionary con-
text. Molecular Ecology 13: 757–773.
Zimmer, E. A., E. H. Roalson, L. E. Skog, J. K. Boggan, and A. Idnurm.
2002. Phylogenetic relationships in the Gesnerioideae (Gesneriaceae)
based on nrDNA ITS and cpDNA trnL–Fand trnE–Tspacer region
sequences. American Journal of Botany 89: 296–311.
APPENDIX 1. Samples used in phylogenetic analyses and their associ-
ated GenBank numbers (ITS/trnL–F/trnE–T) with genera and species
listed alphabetically. All voucher information for these specimens is pre-
viously published in Zimmer et al. (2002) and Roalson et al. (2005a), but
here follow the classification of Roalson et al. (2005b).
Achimenes candida Lindl., AY047065 / AY047124 / AY047183; A. cet-
toana H.E.Moore, AY047066 / AY047125 / AY047184; A. misera Lindl.,
AY047067 / AY047126 / AY047185; Aeschynanthus hildebrandii Hemsl.,
AY047040 / AY047099 / AY047158; Alloplectus herthae Mansf. (previously
identified as Alloplectus bolivianus (Britton) Wiehler), AY047097 /
AY047156 / AY047215; Bellonia spinosa Swartz, AY702350 /AY702394 /
——;Besleria labiosa Hanst., AY047041 / AY047100 / AY047159; Chryso-
themis pulchella (Donn ex Sims) Decne., AY047085 / AY047144 /
AY047203; Codonanthe carnosa (Gardner) Hanst., AY047088 / AY047147 /
AY047206; Columnea spathulata Mansf., AY047092 / AY047151 /
AY047210; Corytoplectus cutucuensis Wiehler, AY047094 / AY047153 /
AY047212; Diastema affine Fritsch, AY702353 / AY702397 / ——;D.
comiferum (DC.) Benth. ex Walp., AY702354 / AY702398 / ——;D. race-
miferum Benth., AY047069 / AY047128 / AY047187; D. scabrum (Poepp.)
Benth. ex Walp., AY702356 / AY702400 / ——;D. vexans H.E.Moore,
AY702357 / AY702401 / ——;Drymonia serrulata (Jacq.) Mart., AY047093
/ AY047152 / AY047211; Episcia lilacina Hanst., AY047091 / AY047150 /
AY047209; E. punctata (Lindl.) Hanst., AY047090 / AY047149 / AY047208;
Eucodonia andrieuxii (DC.) Wiehler, AY047060 / AY047119 / AY047178; E.
verticillata (M.Martens & Galeotti) Wiehler, AY047061 / AY047120 /
AY047179; Gasteranthus quitensis Benth., AY047042 / AY047101 /
AY047160; Gesneria acaulis L., AY047045 / AY047104 / AY047163; G.
christii Urb., AY047046 / AY047105 / AY047164; G. citrina Urb., AY047054
/ AY047113 / AY047172; G. cuneifolia (DC.) Fritsch, AY047047 /
AY047106 / AY047165; G. humilis L., AY047051 / AY047110 / AY047169;
G. pedicellaris Alain, AY047049 / AY047108 / AY047167; G. pedunculosa
(DC.) Fritsch, AY047052 / AY047111 / AY047170; G. reticulata (Griseb.)
Urb., AY047048 / AY047107 / AY047166; G. ventricosa Sw., AY047053 /
AY047112 / AY047171; G. viridiflora (Decne.) Kuntze ssp. sintenisii (Urb.)
L.E.Skog, AY047050 / AY047109 / AY047168; Gloxinia erinoides (DC.)
E.H.Roalson & J.K.Boggan, AY047073 / AY047132 / AY047191; G. peren-
nis (L.) Fritsch, AY047071 / AY047130 / AY047189; G. xanthophylla (Po-
epp.) E.H.Roalson & J.K.Boggan, AY047074 / AY047133 / AY047192;
Gloxinella lindeniana (Regel) E.H.Roalson & J.K.Boggan, AY702361 /
AY702405 / ——;Gloxiniopsis racemosa (Benth.) E.H.Roalson & J.K.Bog-
gan, AY702364 / AY702407 / ——;Goyazia rupicola Taubert, AY702366 /
AY702409 / ——;Heppiella ulmifolia (Kunth) Hanst., AY702369 /
AY702412 / ——;H. viscida (Lindl. & Paxt.) Fritsch, AY702370 /
AY702413 / ——;Kohleria affinis (Fritsch) E.H.Roalson & J.K.Boggan,
AY702351 / AY702395 / ——;K. allenii Standl. & L.O.Wms., AY702371 /
AY702414 / ——;K. amabilis (Planch. & Linden) Fritsch, AY702372 /
AY702415 / ——;K. grandiflora L.P.Kvist & L.E.Skog, AY702373 /
AY702416 / ——;K. hirsuta (Kunth) Regel, AY702374 / AY702417 / ——;
K. peruviana Fritsch, AY702375 / AY702418 / ——;K. rugata (Scheidw.)
L.P.Kvist & L.E.Skog, AY047075 / AY047134 / AY047193; K. sp. nov.
[2446], AY702376 / AY702419 / ——;K. tigridia (J.H.Ohlend.) E.H.Roal-
son & J.K.Boggan, AY702352 / AY702396 / ——;K. trianae (Regel) Hanst.,
AY702377 / AY702420 / ——;K. villosa (Fritsch) Wiehler, AY047076 /
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 173

AY047135 / AY047194; K. warszewiczii (Regel) Hanst., AY702379 /
AY702422 / ——;Lembocarpus amoenus Leeuwenb., AY702380 /
AY702423 / ——;Mandirola sp. aff. ichthyostoma (Gardner) B.C.Seem. ex
Hanst., AY702360 / AY702404 / ——;M. multiflora (Gardner) Decne.,
AY702363 / —— /——;Monopyle flava L.E.Skog, AY702381 / AY702424
/——;M. puberula C.V.Morton, AY047070 / AY047129 / AY047188;
Moussonia deppeana (Schlechtend. & Cham.) Hanst., AY702383 /
AY702426 / ——;M. elegans Decne., AY702384 / AY702427 / ——;M.
septentrionalis (Denham) Wiehler, AY047068 / AY047127 / AY047186;
Napeanthus jelskyi Fritsch, AY047044 / AY047103 / AY047162; Nautiloca-
lyx melittifolius (L.) Wiehler, AY047086 / AY047145 / AY047204; Nema-
tanthus strigillosus (Mart.) H.E.Moore, AY047089 / AY047148 / AY047207;
Neomortonia nummularia (Hanst.) Wiehler, AY047096 / AY047155 /
AY047214; N. rosea Wiehler, AY047095 / AY047154 / AY047213; Niphaea
oblonga Lindl., AY047064 / AY047123 / AY047182; Nomopyle dodsonii
(Wiehler) E.H.Roalson & J.K.Boggan, AY702358 / AY702402 / ——;Pa-
liavana prasinata (Ker Gawl.) Fritsch, AY047081 / AY047140 / AY047199;
Paradrymonia binata Wiehler, AY047087 / AY047146 / AY047205; Pearcea
abunda (Wiehler) L.P.Kvist & L.E.Skog, AY047077 / AY047136 /
AY047195; P. hypocyrtiflora (Hook.f.) Regel, AY702385 / AY702428 / ——;
P. reticulata (Fritsch) L.P.Kvist & L.E.Skog, AY702386 / AY702429 / ——;
P. sprucei (Britton) L.P.Kvist & L.E.Skog, AY702387 / AY702430 / ——;
Pheidonocarpa corymbosa (Swartz) L.E.Skog, AY702388 / AY702431 / ——;
Phinaea albolineata (Hook.) Benth. ex Hemsl., AY702389 / AY702432 /
——;P. divaricata (Poepp.) Wiehler (previously published as P. ecuadorana
Wiehler, which is now considered a synonym of P.divaricata), AY047078
/ AY047137 / AY047196; P. multiflora C.V.Morton, AY702390 / AY702433
/——;Reldia minutiflora (L.E.Skog) L.P.Kvist & L.E.Skog var. minutiflora,
AY047043 / AY047102 / AY047161; Rhytidophyllum auriculatum Hook.,
AY047058 / AY047117 / AY047176; R. exsertum Griseb., AY047055 /
AY047114 / AY047173; R. rupincola (C.Wright) C.V.Morton, AY047057 /
AY047116 / AY047175; R. tomentosum (L.) Mart., AY047056 / AY047115 /
AY047174; R. vernicosum Urb. & Ekman, AY047059 / AY047118 /
AY047177; Seemannia gymnostoma (Griseb.) M.Tours., AY702359 /
AY702403 / ——;S. nematanthodes (Kuntze) J.Schum., AY702362 /
AY702406 / ——;S. purpurascens Rusby [B], AY047072 / AY047131 /
AY047190; S. sylvatica (Kunth) Hanst., AY702365 / AY702408 / ——;
Sinningia cooperi (Paxton) Wiehler, AY047082 / AY047141 / AY047200; S.
incarnata (Aubl.) Denham, AY047083 / AY047142 / AY047201; S. lindleyi
Schauer, AY047084 / AY047143 / AY047202; Smithiantha aurantiaca
Wiehler, AY047063 / AY047122 / AY047181; S. canarina Wiehler,
AY047062 / AY047121 / AY047180; Solenophora calycosa J.D.Sm.,
AY702392 / AY702435 / ——;S. tuerkheimiana J.D.Sm., AY702393 /
AY702436 / ——;Sphaerorrhiza sarmentiana (Gardner ex Hook.) E.H.Roal-
son & J.K.Boggan, AY047079 / AY047138 / AY047197; Streptocarpus
primulifolius Gand., AY047039 / AY047098 / AY047157; Vanhouttea lanata
Fritsch, AY047080 / AY047139 / AY047198.
APPENDIX 2. Gesnerioideae biogeographic distributions used in cod-
ing Fitch Parsimony Optimization (FPO) and DIVA data matrices. Geo-
graphic abbreviations are as follows: C. Am. = Central America; C. I. =
Caribbean Islands; A./W. S. Am. = Andean/Western South America;
S. B. = Southern Brazil; G. = Guianas; and O. W. = Old World.
Species Geographic Distribution
Achimenes candida C. Am.
Achimenes cettoana C. Am.
Achimenes misera C. Am.
Aeschynanthus hildebrandii O. W.
Alloplectus herthae A./W. S. Am.
Besleria labiosa A./W. S. Am.
Chrysothemis pulchella G.
Codonanthe carnosa S. B.
Columnea spathulata A./W. S. Am.
Corytoplectus cutucuensis A./W. S. Am.
Diastema racemiferum A./W. S. Am.
Drymonia serrulata G.
Episcia lilacina C. Am.
Episcia punctata C. Am.
Eucodonia andrieuxii C. Am.
Eucodonia verticillata C. Am.
Gasteranthus quitensis A./W. S. Am.
Gesneria acaulis C. I.
Gesneria christii C. I.
APPENDIX 2. Continued.
Species Geographic Distribution
Gesneria citrina C. I.
Gesneria cuneifolia C. I.
Gesneria humilis C. I.
Gesneria pedicellaris C. I.
Gesneria pedunculosa C. I.
Gesneria reticulata C. I.
Gesneria ventricosa C. I.
Gesneria viridiflora C. I.
Gloxinia erinoides A./W. S. Am.
Gloxinia perennis A./W. S. Am.
Gloxinia xanthophylla A./W. S. Am.
Kohleria rugata C. Am.
Kohleria villosa A./W. S. Am.
Lembocarpus amoenus G.
Monopyle puberula C. Am.
Moussonia septentrionalis C. Am.
Napeanthus jelskyi G.
Nautilocalyx melittifolius C. I.
Nematanthus strigillosus S. B.
Neomortonia nummularia A./W. S. Am.
Neomortonia rosea A./W. S. Am.
Niphaea oblonga C. Am.
Paliavana prasinata S. B.
Paradrymonia binata A./W. S. Am.
Pearcea abunda A./W. S. Am.
Phinaea divaricata A./W. S. Am.
Reldia minutiflora A./W. S. Am.
Rhytidophyllum auriculatum C. I.
Rhytidophyllum exsertum C. I.
Rhytidophyllum rupincola C. I.
Rhytidophyllum tomentosum C. I.
Rhytidophyllum vernicosum C. I.
Seemannia purpurascens A./W. S. Am.
Sinningia cooperi A./W. S. Am.
Sinningia incarnata S. B.
Sinningia lindleyi S. B.
Smithiantha aurantiaca C. Am.
Smithiantha canarina C. Am.
Sphaerorrhiza sarmentiana S. B.
Streptocarpus primulifolius O. W.
Vanhouttea lanata S. B.
APPENDIX 3. Gloxinieae biogeographic distributions used in coding
Fitch Parsimony Optimization (FPO) and DIVA data matrices. Geo-
graphic abbreviations are as follows: C. Am. = Central America; C.I. =
Caribbean Islands; A./W. S. Am. = Andean/Western South America; and
S. B. = Southern Brazil.
Species Geographic Distribution
Achimenes candida C. Am.
Achimenes cettoana C. Am.
Achimenes misera C. Am.
Bellonia spinosa C.I.
Diastema affine A./W. S. Am.
Diastema comiferum A./W. S. Am.
Diastema racemiferum A./W. S. Am.
Diastema scabrum A./W. S. Am.
Diastema vexans A./W. S. Am.
Eucodonia andrieuxii C. Am.
Eucodonia verticillata C. Am.
Gesneria acaulis C.I.
Gesneria pedunculosa C.I.
Gloxinia erinoides A./W. S. Am.
Gloxinia perennis A./W. S. Am.
Gloxinia xanthophylla A./W. S. Am.
Gloxinella lindeniana A./W. S. Am.
Gloxiniopsis racemosa A./W. S. Am.
174 SYSTEMATIC BOTANY [Volume 33

APPENDIX 3. Continued.
Species Geographic Distribution
Goyazia rupicola S. B.
Heppiella ulmifolia A./W. S. Am.
Heppiella viscida A./W. S. Am.
Kohleria affinis A./W. S. Am.
Kohleria tigridia A./W. S. Am.
Kohleria allenii C. Am.
Kohleria amabilis A./W. S. Am.
Kohleria grandiflora A./W. S. Am.
Kohleria hirsuta A./W. S. Am.
Kohleria peruviana A./W. S. Am.
Kohleria rugata C. Am.
Kohleria sp. nov. A./W. S. Am.
Kohleria trianae A./W. S. Am.
Kohleria villosa A./W. S. Am.
Kohleria warszewiczii A./W. S. Am.
Mandirola ichthyostoma S. B.
Mandirola multiflora S. B.
Monopyle flava A./W. S. Am.
Monopyle puberula C. Am.
Moussonia deppeana C. Am.
Moussonia elegans C. Am.
APPENDIX 3. Continued.
Species Geographic Distribution
Moussonia septentrionalis C. Am.
Niphaea oblonga C. Am.
Nomopyle dodsonii A./W. S. Am.
Pearcea abunda A./W. S. Am.
Pearcea hypocyrtiflora A./W. S. Am.
Pearcea reticulata A./W. S. Am.
Pearcea sprucei A./W. S. Am.
Pheidonocarpa corymbosa C.I.
Phinaea albolineata A./W. S. Am.
Phinaea divaricata A./W. S. Am.
Phinaea multiflora C. Am.
Rhytidophyllum auriculatum C.I.
Rhytidophyllum exsertum C.I.
Seemannia gymnostoma A./W. S. Am.
Seemannia nematanthoides A./W. S. Am.
Seemannia purpurascens A./W. S. Am.
Seemannia sylvatica A./W. S. Am.
Smithiantha aurantiaca C. Am.
Smithiantha canarina C. Am.
Solenophora calycosa C. Am.
Solenophora tuerckheimiana C. Am.
2008] ROALSON ET AL.: GLOXINIEAE BIOGEOGRAPHY 175